Akademik Veri Yönetim Sistemi
2016 IEEE Symposium Series on Computational Intelligence, SSCI 2016, Athens, Yunanistan, 6 - 09 Aralık 2016, (Tam Metin Bildiri)
Smoking tobacco products has become a deadly and prevalent habit. It is a known fact that smoking negatively affects the health, economic expenditures, and social life of not only users but also second hand smokers (people's exposure to smoke). Inhaling tobacco smoke can make people vulnerable to nicotine addiction. Correspondingly, second hand smokers may become daily or less than daily smokers in time. The main objective of the presented paper was to classify smoking status of people considering second-hand smoking associated attributes: allowance to second hand smoking at home, allowance to second hand smoking at work, exposure to smoke at work in the past 30 days, beliefs that second hand smoking causes serious diseases, and gender. The classes of smoking status were defined as daily smoker, less than daily smoker and no smoker (not at all). The classification was performed using multilayer perceptron that is a well known neural network approach. The results showed that multilayer perceptron could correctly classify the smoking status of people over % 68 using five attributes. The systems like tobacco use consist of many different interacting factors that make them more complex and dynamic to analyze (classify). For that reason, the % 68 accuracy level can be interpreted as sufficient to the analysis performed for this kind of problems.